Project Concept
To train the Stable Diffusion XL neural network (using the DreamBooth + LoRA method) on my own photographs from a recent trip to Morocco, shot with a compact point-and-shoot camera using a bright flash. All selected images for training share a consistent visual style: heightened saturation, high contrast, and harsh flash lighting.
My photos from the trip 1/ The Atlantic Ocean 2/ Medina, Fez
Project Goal
The main goal of the project was to have the neural network reproduce images in this specific «handwriting» style so that it could generate stylistically similar «vacation photos» from countries I am about to visit for my bucket list.
Photos from the trip
Image Generation: Model and Dataset
For the supplemental work, I used 33 of my own vacation photographs, cropping them to a 1:1 format.
